Make robot activity with intelligent motion. We want to make your robot smarter. You can use our skill & passion! Discover more
01 /  
SIMULATOR

In order to control the movement of the robot according to a predetermined purpose, the process of assigning the appropriate formula and numerical value and measuring the result is repeated many times. This causes wear of the main parts and power consumption and it is also possible due to erroneous input values to discard the parts. But most of all, the biggest loss is the time required for this process. If the software, not the hardware, can learn and optimize the robots’ behavior and monitor the process, the costs including the development period can be greatly reduced.

If the initial robot development was to directly control the hardware, to exchange the motor and gear and to decide the motion, we now use the simulator to infer the data close to the target value, and based on this it is possible to reduce parts and power consumption.

  •   Java-based Rich Client Platform (RCP)
  •   Definition of robot.
  •   Definition of the world.
  •   Behavior Definition.
picture
picture
02 /  
DEEP LEARNING

We would like to find a way to deal with a lot of the costs during the development stage and also to process the important 'motion control' process efficiently. Careful observation reveals that the evolution of robot development is similar to the evolution of artificial intelligence. Initially artificial intelligence utilized the learning method of injecting the rules into the computer, and since the mid-90s, as the Internet has become able to collect vast amounts of data, it has evolved into a form of so-called machine learning in which a large amount of big data is analyzed and the artificial intelligence system learns by itself. But there is a limit to understanding the value of data, however Deep learning beyond this was devised and much has been resolved.

Deep Learning, Reinforcement Learning

First, the robot learns the ability of the elements that make up the robot, and then learns the process of performing the task secondarily. Developers are provided with log and graphical information to monitor this process.

Run learning

The user simply clicks the 'Run learning' button after setting the robot's component and target values, and Simulator leads the automation of the robot development.

03 / API &
ROBOT OS

While learning progresses or after the completion of learning, by providing an API or ROS package it is possible to continue learning with an actual robot. Create an API to communicate the learned control program to the web. Create a package compatible with ROS (Robot Operating System). Extract the most efficient data from the assignment and generate it as a file.

It provides various results depending on application purpose.

  •   Create an API to communicate the learned control program to the web.
  •   Create a package compatible with ROS (Robot Operating System).
  •   Extract the most efficient data from the assignment and generate it as a file.
picture

"Keep on beginning and failing. Each time you fail, start all over again, and you will grow stronger until you have accomplished a purpose - not the one you began with perhaps, but one you'll be glad to remember." - Johanna Mansfield Sullivan Macy -

Why do you choose macroact?

In the entire robot development process, motion control takes up a large part and costs a lot of money. Currently, the appearance of simulators has made it possible to reduce material costs, but it is a fact that we are still investing a lot of time. And it is hard to say if the result is correct because there are too many possibilities in a real environment. Like the victory of the Artificial Intelligence over his human opponent in Go (Baduk), it is very pleasing for a robot manufacturer if you solve the motion control process by introducing the calculation power of the machine.

Definition of robot and world

The most commonly used motors and sensors are pre-registered and listed, making it easy to select, otherwise, you can register it on your own list. You can set the space in which the robot works, the walls, obstacles, and so on.

Behavior Definition

Set the task to be performed by the robot. If you have arms or wheels (legs) - fixed or mobile - for different purposes, you can enter assignments for each part. If there is no input value for each part even if arms and legs are present, the robot learns to recognize the characteristics of each part and utilize it.

Deep Learning, Reinforcement Learning

First, the robot learns the ability of the elements that make up the robot, and then learns the process of performing the task secondarily. Developers are provided with log and graphical information to monitor this process.

Output

It provides various results depending on application purpose. Create an API to communicate the learned control program to the web. Create a package compatible with ROS (Robot Operating System). Extract the most efficient data from the assignment and generate it as a file.

PLEASE JOIN OUR MAILING LIST TODAY

We promise we don't spam and won't give this address away.